Reflections on the use of Bayesian belief networks for adaptive management

A broad range of tools are available for integrated water resource management (IWRM). In the EU research project NeWater, a hypothesis exists that IWRM cannot be realised unless current management regimes undergo a transition toward adaptive management (AM). This includes a structured process of learning, dealing with complexity, uncertainty etc. We assume that it is no longer enough for managers and tool researchers to understand the complexity and uncertainty of the outer natural system-the environment. It is just as important, to understand what goes on in the complex and uncertain participatory processes between the water managers, different stakeholders, authorities and researchers when a specific tool and process is used for environmental management. The paper revisits a case study carried out 2001-2004 where the tool Bayesian networks (BNs) was tested for groundwater management with full stakeholder involvement. With the participation of two researchers (the authors) and two water managers previously involved in the case study, a qualitative interview was prepared and carried out in June 2006. The aim of this ex-post evaluation was to capture and explore the water managers' experience with Bayesian belief networks when used for integrated and adaptive water management and provide a narrative approach for tool enhancement.     

人工神经网络在预测深基坑周边地表沉降变形中的应用研究

深基坑开挖引起的周边地表变形预测是一个复杂非线性问题,引起地表沉降的影响因素很多,各因素之间呈高度的非线性关系。传统的基坑用边地表沉降变形预测方法存在着一定的局限性,其预测精度有待提高,而人工神经网络是一种多元非线性动力学系统,可以灵活方便地对多成因的复杂未知系统进行高度建模,实现全面考虑各种主要影响因素的深基坑周边地表沉降变形预测。本文介绍了误差反向传播(BP)网络模型的结构、学习过程及其算法的改进,径向基函数(RBF)网络模型的结构及其学习过程;分析了影响深基坑开挖周边土体沉降变形的主要影响因素;以25个基坑工程的地表沉降实测资料为训练样本,建立了11个输入影响因素的BP神经网络模型和RBF神经网络模型,通过对样本的学习训练过程及对5个检验样本的预测精度,说明了人工神经网络用于预测基坑周边地表沉降的可行性和准确性。     

Phosphate recovery with granular acid-activated neutralized red mud: Fixed-bed column performance and breakthrough curve modelling

Granular acid-activated neutralized red mud (AaN-RM) has been successfully prepared with good chemical stability and physical strength. However, its potential for industrial application remains unknown. Therefore, the performance of granular AaN-RM for phosphate recovery in a fixed-bed column was investigated. The results demonstrated that the phosphate adsorption performance of granular AaN-RM in a fixed-bed column was affected by various operational parameters, such as the bed depth, flow rate, initial solution pH and initial phosphate concentration. With the optimal empty-bed contact time (EBCT) of 24.27 min, the number of processed bed volumes and the phosphate adsorption capacity reached 496.95 and 84.80 mg/g, respectively. Then, the saturated fixed-bed column could be effectively regenerated with a 0.5 mol/L HCl solution. The desorption efficiency remained as high as 83.45% with a low weight loss of 3.57% in the fifth regeneration cycle. In addition, breakthrough curve modelling showed that a 5-9-1 feed-forward artificial neural network (ANN) could be effectively applied for the optimization of the fixed-bed adsorption system; the coefficient of determination (R²) and the root mean square error (RMSE) evaluated on the validation-testing data were 0.9987 and 0.0183, respectively. Therefore, granular AaN-RM fixed-bed adsorption exhibits promising potential for phosphate removal and recovery from polluted water.     

Particulate matter exchange between atmosphere and roads surfaces in urban areas

The deposition and the re-suspension of particulate matter (PM) in urban areas are the key processes that contribute not only to stormwater pollution, but also to air pollution. However, investigation of the deposition and the re-suspension of PM is challenging because of the difficulties in distinguishing between the resuspended and the deposited PM. This study created two Bayesian Networks (BN) models to explore the deposition and the re-suspension of PM as well as the important influential factors. The outcomes of BN modelling revealed that deposition and re-suspension of PM10 occurred under both, high-traffic and low-traffic conditions, and the re-suspension of PM2.5 occurred under low-traffic conditions. The deposition of PM10 under low-volume traffic condition is 1.6 times higher than under high-volume traffic condition, which is attributed to the decrease in PM10 caused by relatively higher turbulence under high-volume traffic conditions. PM10 is more easily resuspended from road surfaces compared to PM2.5 as the particles which larger than the thickness of the laminar airflow over the road surface are more easily removed from road surfaces. The increase in wind speed contributes to the increase in PM build-up by transporting particulates from roadside areas to the road surfaces and the airborne PM2.5 and PM10 increases with the increase in relative humidity. The study outcomes provide a step improvement in the understanding of the transfer processes of PM2.5 and PM10 between atmosphere and urban road surfaces, which in turn will contribute to the effective design of mitigation measures for urban stormwater and air pollution.     

基于BP神经网络的北京城区公园土壤PAHs含量预测

城市公园是城市生态环境的重要组成部分,其环境质量与人类健康息息相关.选择北京市121个城区公园,采集公园土壤样品并分析其中7种多环芳烃（PAHs）含量,评价城区公园土壤中PAHs的含量水平,并基于BP神经网络预测了2020年和2023年土壤PAHs含量.结果表明:北京城区公园土壤中w（PAHs）（7种PAHs总含量）范围为0.033~4.182 mg/kg,低于GB 36600—2018《土壤环境质量建设用地土壤污染风险管控标准（试行）》土壤污染风险筛选值,且7种PAHs的毒性当量浓度（TEQ）均低于世界卫生组织标准值（1 mg/kg）,对人体健康的毒性风险较小.将14个影响指标（8个社会经济因子与6个公园特征因子）作为输入层、土壤w（PAHs）作为输出层,建立BP神经网络的拟合优度达0.845.预测结果显示,2020年和2023年北京城区公园土壤中w（PAHs）范围分别为0.008~0.969 mg/kg和0.022~1.988 mg/kg,整体均低于GB 36600—2018土壤污染风险筛选值,但随时间推移呈上升趋势,尤其朝阳区和海淀区将有大幅增长.研究显示:城市化发展因素对土壤w（PAHs）的增加有明显影响,城市发展进程影响不容忽视;至2023年,北京城区公园土壤若不加管理,其w（PAHs）将持续增长.      

基于BP神经网络的三峡库区土壤侵蚀强度模拟

降雨侵蚀力变化是一复杂过程,其变化存在一定的随机波动性,土壤侵蚀是三峡库区生态环境脆弱最主要的影响因素之一,查明库区土壤侵蚀强度的演化过程及未来趋势是库区生态文明建设过程中急需解决的关键科学问题。论文基于三峡库区1990年侵蚀降雨特征,利用BP神经网络对2010年75个站点降雨侵蚀力进行模拟、验证,预测2030年75个站点降雨侵蚀力。选取2030年预测结果中位于库区周围的27个站点,结合2030年库区自然增长、生态保护情景下土地利用模拟数据,使用RUSLE计算2030年土壤侵蚀强度。结果表明：1）2010年库区降雨侵蚀力模拟相对误差为15%,测试样本数据相对误差为14.67%,预测相对误差为19.65%,NE系数为0.85,说明BP神经网络对库区降雨侵蚀力具有良好模拟效果;2）2010年库区土壤侵蚀强度的Kappa指数为0.75,计算结果能满足模拟与预测需求;3）在土地利用不变情况下,2030年库区轻度、中度侵蚀面积均有所增加,微度及强烈以上侵蚀面积均呈减少趋势,且侵蚀强度转变中的58%来源于相邻侵蚀强度,跨侵蚀等级区的较少;4）在降雨侵蚀力不变情况下,自然增长、生态保护情景下未来土地利用变化所导致的土壤侵蚀均呈下降趋势,后者下降的趋势更为明显;5）在降雨侵蚀力及土地利用均变化的情况下,自然增长、生态保护情景下土壤侵蚀均呈下降趋势。     

基于SOM-SVM组合分类器的涂层防护性能研究

目的为避免EIS,EN技术可能出现的问题,建立一个准确、高效的评价模型,以探究现役军用有机涂层防护性能。方法利用电化学阻抗谱(EIS)、电化学噪声(EN)技术分析了两种军车有机涂层在循环暴露试验中的腐蚀行为,提取低频阻抗模值|Z|_(0.1 Hz)与涂层噪声电阻R_n两种电化学评价参数作为自组织神经网络(SOM)的输入训练样本,同时结合支持向量机(SVM)方法建立涂层防护性能组合分类器。结果将涂层失效过程自适应地分为涂层防护性能良好、防护性能下降、基本失效三个阶段。结论所建立的SOM-SVM组合分类器对于辅助分析涂层防护性能具有可行性。     

铝合金的实验室盐雾试验腐蚀行为图像特征提取

目的提取铝合金材料在盐雾箱试验中可以表征腐蚀程度的表面形貌图像特征量。方法首先,对采集于表面化学清洗过的试样原始图像进行图像增强等预处理,突出腐蚀部位;其次,基于数字图像处理的分形和小波分解方法提取出分形维数和小波能量特征值。结果与以质量损失量为基础的腐蚀深度特征值进行相关性对比,基于图像分析的特征提取法的准确度和精度比较高。结论该方法可以应用于对试样的腐蚀程度进行定性和定量分析,判断并预测试样的腐蚀速度。     

Analysis of pollutant levels in central Hong Kong applying neural network method with particle swarm optimization

Air pollution has emerged as an imminent issue in modernsociety. Prediction of pollutant levels is an importantresearch topic in atmospheric environment today. For fulfillingsuch prediction, the use of neural network (NN), and inparticular the multi-layer perceptrons, has presented to be acost-effective technique superior to traditional statisticalmethods. But their training, usually with back-propagation (BP)algorithm or other gradient algorithms, is often with certaindrawbacks, such as: 1) very slow convergence, and 2) easilygetting stuck in a local minimum. In this paper, a newlydeveloped method, particle swarm optimization (PSO) model, isadopted to train perceptrons, to predict pollutant levels, andas a result, a PSO-based neural network approach is presented. The approach is demonstrated to be feasible and effective bypredicting some real air-quality problems.     

Using improved neural network model to analyze RSP,NOx and NO2 levels in urban air in Mong Kok,Hong Kong

As the health impact of air pollutants existing in ambient addresses much attention in recent years, forecasting of airpollutant parameters becomes an important and popular topic inenvironmental science. Airborne pollution is a serious, and willbe a major problem in Hong Kong within the next few years. InHong Kong, Respirable Suspended Particulate (RSP) and NitrogenOxides NO_x and NO₂ are major air pollutants due to thedominant diesel fuel usage by public transportation and heavyvehicles. Hence, the investigation and prediction of the influence and the tendency of these pollutants are ofsignificance to public and the city image. The multi-layerperceptron (MLP) neural network is regarded as a reliable andcost-effective method to achieve such tasks. The works presentedhere involve developing an improved neural network model, whichcombines the principal component analysis (PCA) technique and theradial basis function (RBF) network, and forecasting thepollutant levels and tendencies based in the recorded data. Inthe study, the PCA is firstly used to reduce and orthogonalizethe original input variables (data), these treated variables arethen used as new input vectors in RBF neural network modelestablished for forecasting the pollutant tendencies. Comparingwith the general neural network models, the proposed modelpossesses simpler network architecture, faster training speed,and more satisfactory predicting performance. This improvedmodel is evaluated by using hourly time series of RSP, NO_x and NO₂ concentrations collected at Mong Kok Roadside Gaseous Monitory Station in Hong Kong during the year 2000. By comparing the predicted RSP, NO_x and NO₂ concentrationswith the actual data of these pollutants recorded at the monitorystation, the effectiveness of the proposed model has been proven.Therefore, in authors' opinion, the model presented in the paper is a potential tool in forecasting air quality parameters and hasadvantages over the traditional neural network methods.